The fluid, seemingly effortless execution of sequences of movements is a ubiquitous feature of everyday motor behavior. Ample evidence for the importance of this ability comes from the common human diseases (Parkinson's disease, in particular) in which sequential skills are especially impaired. Long-term motor sequencing skills are formed, most likely, through the cooperation of parallel cortical-sub-cortical circuits involving associative, premotor, and motor regions of the brain. Recent evidence suggests that for each of these brain circuits, the sub-cortical loop through the basal ganglia (BG) contributes selectively to reinforcement-driven modulation of thalamo-cortical plasticity while cortex is well-suited as a site for long-term retention and efficient recall of well-practiced skills. These findings lead to the hypotheses that BG loops play central roles in the acquisition of sequence information whereas the anatomically-connected cortical areas are more important for the storage and use of already-learned information. The specific aims (SAs) of this proposal will test that general hypothesis by using non-human primates: (1) To determine if neurons in the globus pallidus interna (GPi, a primary BG output nucleus) preferentially encode sequence task information during learning and the production of recently learned sequences. Cortical neurons are predicted to not show a preference for recently learned sequences. (2) To test if intact BG circuits are necessary primarily for the learning and production of recently- learned sequences. Cortical circuits, in contrast, are predicted to be necessary even for well-learned sequences. Associative loops through cortex and BG may play greater roles in the fast acquisition and flexible recall of goal- directed sequence information. The premotor and motor loop circuits may mediate slow acquisition of habit-like effector-specific representations. We will infer the circuit membership of individual GPi neurons by stimulating different cortical areas and observing the orthodromic inhibitory effects. Animals will perform a discrete sequence production task alternating in blocks between random, novel-to-familiar and over-trained sequences. SA1 will test if neuronal encoding of task information in associative, premotor, and motor circuits reflects the predicted divergent roles for BG- and cortical-components of these circuits. SA2 will determine if interruptions of BG output (i.e., GPi inactivation or lesion) selectively impair the learning or recall of recently learned sequences. Inactivations of cortex, in contrast, are predicted to also disrupt the recall of well-learned sequences. Results from these experiments will aid in understanding the physiological basis for normal and impaired sequential behavior in humans.